Topics
The Topics tab shows which broad subject areas appear in the posts matching your query.
The Topics tab of the Social Intelligence dashboard counts matching posts by subject area, such as News, Technology, or Culture. Its results show the topic counts and proportions for the submitted query, along with their changes across the selected time range.
The topic mix provides context for the other tabs. A negative sentiment score could accompany news coverage, personal accounts, or discussion of a film. Seeing which categories appear helps you understand what kinds of posts contributed to the score before examining their sentiment or emotion labels in more detail.
A post can have several topic labels. A post discussing a film's financial results, for example, could be labeled Culture and Business. Both labels describe the same post, so adding the category counts can produce a total larger than the number of posts. The charts use these counts in different ways.
Topic categories
Topics use a fixed set of broad categories. They organize the text selected by your query; the individual events, people, and issues being discussed remain in the posts themselves.
| Category | Subject area |
|---|---|
| News | News, public affairs, and social issues |
| Business | Business activity, companies, and entrepreneurship |
| Technology | Science and technology |
| Sports | Sporting events, teams, and competition |
| Culture | Film, television, music, gaming, celebrity, arts, and education |
| Lifestyle | Hobbies, food, fitness, health, travel, and fashion |
| Daily Life | Personal updates, relationships, family, and student life |
| Misc | Posts for which the classifier has not assigned another subject category |
These are classifications of the post's text. A category does not identify the author's occupation or audience. Misc can contain relevant posts whose subject the classifier did not place in a named category. The topic badges in the Recent posts feed show the labels assigned to each post, allowing you to compare the classification with the text.
Reading one query
The single-query layout also appears when you choose one query through Select or combine several through Aggregate. Getting Started explains those controls.
Topics volume
Topics volume shows a bar for each category. The colored segment counts posts with that label, and the full track, including its gray remainder, represents all matching posts. Hovering over a bar shows the category count and the total post count. A post contributes at most once to each category's bar.
Topics distribution
The Topics distribution pie divides each category's count by the sum of all topic counts in the chart. Each label counted for a post is one topic label assignment. The pie shows the share of these assignments, which can differ from the share of posts.
Suppose 100 posts have 150 topic label assignments between them, and 60 posts carry Technology. The Technology bar fills 60% of its track. Its pie slice is 40%, because 60 of the 150 assignments are Technology. The same denominator is used by the topic percentages in the stats bar.
Stats bar
The stats bar summarizes the category shares over the selected range and highlights the interval with the strongest concentration.
| Stat | What it measures |
|---|---|
| Top topic | The category with the largest count and its share of the topic label assignments. |
| Runner-up topic | The category with the second-largest count and its label share. |
| Topic diversity | How many categories each account for at least 10% of the topic label assignments. One is labeled Concentrated, two or three Moderate, and four or more Diverse. |
| Highest concentration | The largest share reached by any topic in a time interval, shown with the topic, its percentage, and the interval's start date. |
Topic diversity summarizes how the labels are distributed. Several categories can come from the same posts and authors, so the count does not establish how many distinct audiences are involved.
Top topic and Runner-up topic give an initial sense of the subject mix, while the pie shows how much the leading categories outweigh the rest. An unexpected leading category can also be a reason to read the matching posts. A product name, for example, may be matching another meaning of the same word. If that is happening, a phrase or an additional term can make the query more specific.
Highest concentration compares the topic shares within individual daily or weekly intervals. The winning category can differ from Top topic, which uses the full range. When several intervals are available, this stat leaves out the final displayed interval.
Topics over time
The chart groups posts into daily or weekly intervals, chosen automatically from the time range. Each color identifies a topic, and dates follow your timezone.
In Volume mode, a segment counts the posts carrying its topic label in that interval. Because a post can appear in several segments, the full stack counts topic label assignments. The Volume tab shows the corresponding post counts without adding their labels together.
In Distribution mode, each interval with topic labels is scaled to 100%. A segment shows its topic's share of the assignments in that interval. A growing share can accompany an unchanged count if the other topic counts fall. Switching between the two modes shows whether a change in proportion also involved a change in the number of posts carrying that label.
The chart includes the final interval even when Highest concentration excludes it. If that interval is still underway, its counts describe only the available portion. Reading the posts from an interval helps identify which discussions contributed to its topic mix.
Combining queries with Aggregate
Aggregate produces a combined result from posts matching any of the queries, counting each shared post once. It then uses the single-query charts and stats described above. A shared post can still contribute several different topic labels within that combined result.
For queries naming a company and its products, this view shows the topic mix across all their matching posts. Compare and Select retain the separate breakdowns for examining which discussions each query contributes.
Comparing queries
With two or more submitted queries, Compare shows their topic distributions together. Each query uses the same time range and topic filter. A post matching several queries contributes its labels to each of them.
Topic distributions and comparison stats
The Topics chart gives each query a bar divided into topic shares. A bar with topic labels adds to 100%, using the sum of that query's topic counts as its denominator. The colors identify topics, allowing you to follow the same category across bars. Equal-length bars can represent very different post counts; the query totals and Volume tab show that difference.
Top topic identifies the largest topic share found within any query and shows the query's name alongside it. Runner-up topic identifies a different category with the next-largest share found across the queries. Each category is considered once, using its highest share. The two stats can therefore name the same query, or different queries.
Highest variation identifies the topic with the largest gap between its highest and lowest shares across queries. The gap is measured in percentage points. For example, Technology shares of 50% and 20% produce a gap of 30 percentage points, displayed in the cell as 30.0%. This compares the queries over the selected range; the time chart below shows when their counts changed.
Most diverse names the query with the most categories at or above the 10% label-share threshold. This is the same count used by Topic diversity in the single-query layout.
Differences in the topic mix can help explain differences seen in the other tabs. Two subjects with similar sentiment scores may have different balances of Business, Culture, and Daily Life posts. Comparing their topic distributions makes that difference visible; Select and the feed then provide the text behind each result.
One topic over time
The comparison's Topics over time chart follows one selected topic across all queries. It initially selects the category with the largest share within any query. The dropdown changes the topic, and the segments use query colors.
In Volume mode, each segment counts that query's posts carrying the selected topic label. In Distribution mode, the denominator is the sum of those counts across queries within the interval. This percentage answers how the selected topic's counts are divided between the queries, rather than how prominent that topic is within each query.
For example, if one query has 60 Technology posts on a day and another has 40, their segments in Distribution mode are 60% and 40%. These percentages do not depend on how many posts in either query carry other topic labels. A post matching both queries is counted in both segments, so the sum can include overlap.
Narrowing the results by topic
The query bar's Filter control selects posts by their assigned topic labels. Apply saves the category selection, and Analyze in Write queries runs the existing queries with that filter. Selecting several categories keeps posts carrying any of them. All and None both remove the topic restriction.
The filter selects whole posts and retains their other labels. If a post has both Technology and Business, selecting Technology keeps the post and both of its labels. The resulting distribution can therefore still contain Business and other categories. The selected topic filter also applies when you open the Sentiment, Emotions, and other tabs.
The feed's separate Filter by sentiment or emotion box narrows the posts already loaded by sentiment or emotion and leaves the charts unchanged. On the Topics tab, the topic badges beneath each post provide the category information for closer reading. Query Syntax explains how text operators and the topic filter work together.